Marvell’s AI connectivity and custom-chip expansion can support substantially higher earnings, but revenue growth alone does not settle the value of each share. Our November 2027 standalone Base scenario is $247.82, using FY2029 earnings after recurring share compensation. That is above the October 30 close but 27.23% below the dated analyst mean. The largest immediate uncertainty is whether a major acquisition changes the financing and ownership assumptions before this growth arrives.
Research date: November 1, 2026. Financial statements are for the quarter ended August 1, 2026, not the just-ended October quarter. FY2029 denotes Marvell’s fiscal year ending around January 2029. The reference regular-session close was $196.40 on October 30 at 4:00 p.m. EDT; the provider reported $172.23 billion market capitalization. This is a large-cap company from our existing tracking universe, not a newly discovered sub-$10 billion stock. We found no dedicated Company Research baseline on the current site, but preserve our earlier archived analysis below.
1. The business: more data movement, not just more accelerators
Marvell supplies the silicon that helps data move within and between computing systems, alongside custom chips designed for particular customers. Its October 6 Investor Day presentation groups the opportunity into interconnect, switching and storage, custom silicon, and communications/other. Digital signal processors handle high-speed signals; switches direct traffic; custom silicon adapts computing or supporting functions to a customer’s architecture. These are complementary opportunities, not identical revenue streams.
The economic attraction is that a larger AI cluster needs more than additional computing chips. It also needs reliable communication among those chips and access to memory and storage. Marvell can earn content from several parts of that system. The limit is that design selection is only the beginning: qualification, production, deployment and collection must follow. A customer can increase its total infrastructure budget without every supplier capturing the same growth.
Standard products can spread engineering investment across customers. Custom programs may offer long-lived relationships but require expensive development before volume, and a few large deployments can dominate the growth path. Competition is therefore about power, performance, reliability, software compatibility and delivery—not simply having an optical product. Customer-designed alternatives and rival suppliers can reduce both prices and the content available to Marvell.
Our recent Credo analysis examines optical expansion against recurring costs, while Astera Labs research explains another part of the connectivity market. Those comparisons frame the industry question; they do not imply identical products, shared customers or interchangeable valuation multiples.
2. A much larger company plan still needs an earnings bridge
| Evidence | Earlier position | Latest confirmed position | What remains unproven |
|---|---|---|---|
| Q2 FY2027 revenue | $2.700bn guidance midpoint | $2.7393bn actual; $39.3m above the rounded midpoint | Whether faster shipments sustain margins and cash conversion |
| FY2028 total revenue outlook | $18bn August outlook | Approximately $20bn at October Investor Day | Customer deployment and product mix |
| FY2029 custom revenue | Prior $10bn target | More than $12bn company target | Production volumes, not merely design wins |
| FY2031 financial model | Long-term ambition | $70–90bn revenue; 44–46% non-GAAP operating margin | Several years of extraordinary growth and operating leverage |
The scale of the new plan changes the question. It is no longer enough to ask whether AI spending is growing. Investors must ask how much of that spending becomes Marvell revenue, how much engineering expense is required to earn it, and how much remains after compensation and financing. The company’s long-term non-GAAP margin is a destination—not an appropriate margin to attach automatically to next year’s sales.
The Celestial AI completion announcement initially placed revenue contributions in the second half of FY2028, reaching a $500 million annualized run rate in that year’s fourth quarter and $1 billion a year later. Those are company expectations and exit rates, not full-year recognized revenue. Optical scale-up expands the opportunity but adds adoption risk; it does not eliminate the need to prove the technology in customer systems.
3. Current profit and cash are not the same as adjusted EPS
| USD millions, unless stated | Q2 FY2027 | First half FY2027 |
|---|---|---|
| Revenue | 2,739.3 | 5,157.1 |
| GAAP net income | 308.0 | 342.5 |
| Stock-based compensation | 326.2 | 533.8 |
| Operating cash flow | 605.5 | 1,244.3 |
| Property/equipment purchases | 126.7 | 282.4 |
| CFO less cash capex: our simple FCF | 478.8 | 961.9 |
| After technology-license purchases | 474.3 | 956.9 |
Q2 GAAP EPS was $0.33 versus $0.94 non-GAAP. A company can have genuine operating momentum while the adjusted measure overstates the earnings attributable to shareholders if recurring costs are ignored. Our first-half cash flow after capex and technology-license purchases falls from $956.9 million to $423.1 million if the reported compensation expense is treated as an illustrative economic charge. That is not GAAP cash flow, nor a claim that the company paid that amount in cash; it tests how much of the apparent cash surplus depends on compensating people with ownership.
This diagnostic also is not a direct stock valuation. Share awards have timing and vesting conditions, and accounting expense is not the same as the market value of all outstanding claims. We therefore keep the operating-cost question separate from the share-count question in the forward model. Nor do we annualize one good cash quarter into a permanent yield.
The 10-Q reports $3,932.8m cash, $4,999.9m debt principal ($4,962.9m carrying amount), and $412.0m first-half receivables factoring. Principal net of cash is $1,067.1m. Customer A supplied 16% of Q2 revenue; Distributor A represented 44%. A distributor is not necessarily one ultimate customer.
Factoring changes collection timing; subtracting every receivable sold would not recreate a clean counterfactual cash flow. Customer concentration likewise needs interpretation: billing-channel concentration and final demand concentration are different. For our thesis, the relevant test is whether growth is accompanied by manageable inventories, collectible receivables and adequate returns on the engineering and supply commitments required to support it.
The October 23 Form 8-K documents a $3 billion revolving facility replacing $1.5 billion, with a conditional additional $2 billion option. No revolving loans were outstanding at its October 22 effective date. Undrawn capacity provides flexibility; it is neither cash already received nor debt already drawn, and borrowing would introduce additional interest and covenants.
4. Google, NVIDIA and acquisition claims: growth has an ownership cost
The August 19 Google agreement filing identifies custom programs attached to Google’s TPU ecosystem and a warrant for up to 58,970,907 shares at $206.58. Of these, 1,360,867 vest over the first year; the rest depend on discretionary purchases in 240 tranches of $500 million through FY2033. Thus $120 billion is the cumulative purchasing threshold for all performance tranches—not committed backlog or present revenue.
This structure couples potential customer success with dilution. It would be misleading to recognize the entire revenue threshold now, but equally misleading to disregard the warrant because it was out of the money at the reference close. If the business succeeds and the share price rises, the claim can matter. Exercise proceeds also matter: adding every warrant share while ignoring its exercise price is a harsh no-proceeds stress, not a neutral diluted-EPS forecast.
NVIDIA’s March partnership links Marvell custom products to NVLink Fusion and included a $2 billion investment. The 10-Q shows 921.2m Q2 diluted weighted-average shares already including the preferred-stock conversion effect. We do not add NVIDIA’s approximately 21.8m conversion shares again. Celestial’s Note 6 lists maximum settlement of $233m cash plus 22.4m shares; the risk section says 24.4m shares. We preserve that discrepancy and reserve the larger share figure rather than inventing a reconciliation.
Our 960 million economic-share assumption starts from the 921.2 million reported diluted average, adds 24.4 million for the larger stated acquisition contingency, and allows 14.4 million for other existing-claim and measurement uncertainty. It is a modeling budget, not a verified future GAAP diluted count or a promise of buybacks. The Google warrant is stress-tested separately below. Recurring future share compensation remains an expense; we do not also compound an unlimited new-grant dilution rate on top of that charge.
5. November 2027 standalone scenarios using FY2029 earnings
Our primary method is forward earnings, because Marvell already generates material revenue and profit. These are AHNIM normalized economic forecasts, not company guidance or GAAP EPS forecasts. Operating margins retain recurring share compensation but exclude legacy acquired-intangible amortization and exceptional acquisition fair-value remeasurement. We assume no new major acquisition, no new exceptional tax gain and no separately credited surplus cash.
| AHNIM FY2029 input / output | Bear | Base | Bull |
|---|---|---|---|
| Revenue ($bn) | 24 | 30 | 36 |
| Custom / other data-center / comms revenue ($bn) | 10 / 12 / 2 | 12.5 / 15.3 / 2.2 | 15 / 18.7 / 2.3 |
| Economic operating margin | 25% | 32% | 37% |
| Net interest/other expense ($m) | 150 | 150 | 150 |
| Normalized tax rate | 16% | 16% | 16% |
| Economic shares before Google stress (m) | 960 | 960 | 960 |
| Calculated economic EPS | $5.11875 | $8.26875 | $11.52375 |
| Forward P/E | 24× | 30× | 35× |
| One-off contingent cash reserve ($m) | 233 | 233 | 233 |
| November 2027 scenario target | $122.61 | $247.82 | $403.09 |
| Versus October 30 close | −37.57% | +26.18% | +105.24% |
| Versus analyst mean $340.56 | −64.00% | −27.23% | +18.36% |
Formula: economic earnings = (revenue × operating margin − net interest/other expense) × (1 − tax); target = (earnings × P/E − contingent cash reserve) ÷ shares. Base: [($30,000m × 32% − $150m) × 84% × 30 − $233m] ÷ 960m = $247.82. The cash reserve is conservative, not a claim payment is certain. We do not also subtract the accounting fair-value liability, which would overlap the modeled cash and share claims.
Our $12.5 billion Base custom assumption is modestly above the $12 billion threshold in the company’s “more than $12 billion” FY2029 target; that open-ended target is not a precise point estimate. The remaining $17.5 billion requires continued interconnect, switching and storage growth alongside a modest communications contribution. Bear assumes delayed custom ramps and lower operating leverage; Bull needs broader adoption and stronger mix. These are explicit business assumptions, not percentages selected to reproduce an analyst target.
A 32% cost-inclusive Base margin is demanding. Q2 non-GAAP operating margin of approximately 36.6%, less compensation equivalent to 11.9% of revenue, yields roughly 24.7% before other accounting differences. Our Base therefore needs about seven additional points of operating leverage. The 24×/30×/35× multiples are judgmental growth premiums, not measured peer averages. Base does not become conservative merely because it is below consensus.
At Base earnings, 25× implies $206.48 and 35× $289.16. An additional $1 billion revenue at the same margin adds $8.40 per share; one margin point adds $7.88. Raising the denominator to one billion shares reduces Base to $237.91. These separate sensitivities should not be added together as though the variables were independent in the real business.
For an illustrative full-Google-vesting stress, we hold operating earnings fixed and include the cash exercise proceeds: stressed price = (pre-warrant equity value + 58.970907m × $206.58) ÷ (960m + 58.970907m), only when exercise is in the money. Base becomes $245.43 and Bull $391.72; Bear remains $122.61 because the exercise price exceeds that scenario. This is not an actual vesting forecast or formal treasury-stock EPS calculation. Full performance vesting requires purchases far beyond our near-term revenue model, so the exercise is a separate ownership stress, not a claim that all shares already exist.
6. A cash-flow check gives a more cautious answer
Our supporting DCF illustrates how much the earnings multiple depends on growth duration. Using a simplified late-2027 valuation origin, we discount five annual cash flows for FY2029–FY2033 at year-end intervals. Revenue is $30/$40/$52/$62/$70 billion; economic operating margins are 32%/33%/34%/35%/35%; tax is 16%; net reinvestment is 5% of revenue. These are independent AHNIM assumptions, not the company’s free-cash-flow guidance.
Cash flow = revenue × margin × 84% − net reinvestment. Terminal growth is 3%, with terminal reinvestment determined by growth divided by an assumed 25% return on invested capital. We deduct $1,067.1 million net debt and the $233 million cash reserve, divide by 960 million shares, and do not add Google exercise proceeds in this base check. Keeping August net debt unchanged is an explicit simplification, not a forecast that intervening cash generation disappears.
This produces $164.00 at a 12% discount rate, $216.44 at 10% and $130.86 at 14%. The 12% rate is our risk assumption, not a measured market WACC. The $247.82 earnings-multiple Base therefore requires a richer growth-duration valuation than this cash-flow path supports. We do not average the two methods or use DCF as an automatic ceiling. The disagreement is a warning: the apparent upside at $196.40 depends materially on the market continuing to pay for expansion beyond near-term cash generation. On a cash-flow-only reading, the margin of safety is weaker.
7. Analyst targets and our earlier model are different comparisons
Stock Analysis’s S&P Global snapshot, updated October 29 and retrieved November 1, reports 47 analysts: one-year mean $340.56, median $350 and range $229.95–450. Our November 2027 horizon is approximately comparable, but constituent cutoff dates and earnings definitions are not identical. Base is 27.23% below the mean, calculated from the published target; the absolute-difference test is |247.82 − 340.56| ÷ 340.56. Bear and Bull comparisons appear separately above.
Our lower Base reflects cost-inclusive margins, an explicit claim reserve and a 30× earnings multiple. At our earnings and cash reserve, the mean would require approximately 41.2× rather than 30×. We cannot allocate the full difference to each analyst’s assumptions without their models. The gap is a reason to review our forecasts and valuation basis—not a reason to force the target toward consensus.
The Fly reported on October 7 that Craig-Hallum raised its target from $300 to $380 and retained Buy following Investor Day, citing stronger AI-silicon and optical demand. We reviewed this secondary account, not the original broker note. It is an earlier target increase, not a reaction to an already-reported October-quarter result or confirmation of a Lumentum transaction.
| Model history | Archived August 20, 2026 analysis | November 1, 2026 analysis |
|---|---|---|
| Earnings period / definition | FY2028 non-GAAP earnings | FY2029 normalized economic earnings, recurring compensation retained |
| Revenue Bear / Base / Bull | $14bn / $16.5bn / $18bn | $24bn / $30bn / $36bn |
| Base margin and denominator | 32% net margin; about 921.36m shares after time-based warrant allowance | 32% operating margin before interest/tax; 960m economic shares plus separate Google stress |
| P/E Bear / Base / Bull | 36× / 42.5× / 50× | 24× / 30× / 35× |
| Published targets Bear / Base / Bull | $170 / $245 / $355 | $122.61 / $247.82 / $403.09 |
The near-unchanged Base target hides a substantial method and horizon change. Higher revenue and a later earnings year are offset by explicit recurring costs, more claims and a lower multiple. This is not a like-for-like statement that our FY2028 EPS estimate rose or fell. We leave the earlier article and its dated assumptions intact rather than rewriting its original conclusion.
8. The acquisition boundary—and the next evidence
Reuters reported on October 28, republished by MarketScreener, that Marvell was in advanced discussions to acquire Lumentum; neither company immediately commented. We have not verified an executed transaction or final financing terms. Our targets exclude Lumentum revenue, synergies, purchase consideration, financing and new shares. A signed deal could make this standalone model obsolete and would trigger a new enterprise-to-equity and ownership calculation.
Until then, the next checks are concrete: reported October-quarter revenue and margins versus the existing $3.150 billion ±5% outlook; progress toward the FY2028 revenue plan; production evidence behind custom and optical ramps; cash conversion after working capital and licensing payments; and the actual diluted-share and warrant disclosures. The quarter has ended, but that does not make its financial results available.
We would raise operating assumptions only when deployment and unit economics support them, and lower them for qualification delays, weaker mix, customer bargaining pressure or capital requirements that rise faster than earnings. A financing announcement warrants a capital-structure revision even if the strategic rationale looks attractive. Our conclusion is therefore conditional: Marvell has a credible route to a much larger business, while its value per share still depends on how expensively that route is built.
November 3, 2026 update: Marvell has now signed the definitive all-stock agreement. Our deal analysis calculates the 67/33 ownership split, the 49.34% reference-share increase and the limited static EPS accretion before synergies and transaction effects.
This article is for information and education, not a personalized recommendation or a guarantee of returns. Our research, AI-use and correction policy explains our responsibility for distinguishing source evidence from assumptions.
Sources and original documents
Q2 FY2027 results; August 27, 2026; quarter ended August 1. https://investor.marvell.com/news-events/press-releases/detail/1031/marvell-technology-inc-reports-second-quarter-of-fiscal-year-2027-financial-results
Form 10-Q; filed August 28, 2026; period ended August 1. https://investor.marvell.com/sec-filings/all-sec-filings/content/0001835632-26-000025/mrvl-20260801.htm
Investor Day presentation; October 6, 2026; printed slide numbers cited above. https://d1io3yog0oux5.cloudfront.net/_2122bc187be9454fb556c7291c93ad17/marvell/files/654831/marvell-investor-day-2026-presentaion-deck-print-version.pdf
Google commercial agreement and warrant Form 8-K; filed August 19, 2026. https://investor.marvell.com/sec-filings/all-sec-filings/content/0001193125-26-356217/d412696d8k.htm
Revolving credit agreement Form 8-K; filed October 23, 2026; effective October 22. https://investor.marvell.com/sec-filings/all-sec-filings/content/0001193125-26-430424/0001193125-26-430424.pdf
Celestial AI acquisition completion; February 2, 2026; dated revenue expectations. https://www.marvell.com/company/newsroom/marvell-completes-acquisition-of-celestial-ai.html
NVIDIA NVLink Fusion partnership and investment; March 31, 2026. https://www.marvell.com/company/newsroom/nvidia-ai-ecosystem-expands-marvell-joins-forces-through-nvlink-fusion.html
MRVL regular-session quote and market capitalization; October 30 close; retrieved November 1. https://stockanalysis.com/stocks/mrvl/
Stock Analysis / S&P Global consensus; 47 analysts; updated October 29, retrieved November 1, 2026. https://stockanalysis.com/stocks/mrvl/forecast/
The Fly / TipRanks report of Craig-Hallum target increase; October 7, 2026; secondary report. https://www.tipranks.com/news/the-fly/marvell-price-target-raised-to-380-from-300-at-craig-hallum-thefly-news
Reuters via MarketScreener; October 28 report, October 29 update; unconfirmed acquisition discussions. https://uk.marketscreener.com/news/marvell-in-advanced-talks-to-buy-lumentum-bloomberg-news-reports-ce785cd3db8af523
Historical AHNIM-related Korean research note; August 20, 2026; checked against user-supplied archive, not a live revalidation. https://blog.naver.com/c4choi_1/224383814744
Leave a Reply